Changeset 58418

Timestamp:

06/15/2024 06:31:24 AM (6 months ago)

Author:

dmsnell

Message:

KSES: Preserve some additional invalid HTML comment syntaxes.

When wp_kses_split processes a document it attempts to leave HTML comments
alone. It makes minor adjustments, but leaves the comments in the document in
its output. Unfortunately it only recognizes one kind of HTML comment and
rejects many others.

This patch makes a minor adjustment to the algorithm in wp_kses_split to
recognize and preserve an additional kind of HTML comment: closing tags with
an invalid tag name, e.g. </%dolly>.

These invalid closing tags must be interpreted as comments by a browser.
This bug fix aligns the implementation of wp_kses_split() more closely
with its stated goal of leaving HTML comments as comments.

It doesn't attempt to fully fix the mis-parsed comments, but it does propose a
minor fix that hopefully won't break any existing code or projects.

Developed in ​https://github.com/WordPress/wordpress-develop/pull/6395
Discussed in https://core.trac.wordpress.org/ticket/61009

Props ellatrix, dmsnell, joemcgill, jorbin, westonruter, zieladam.
See #61009.

Location:

trunk

Files:

• src/wp-includes/kses.php (modified) (5 diffs)
• tests/phpunit/tests/kses.php (modified) (1 diff)

trunk/src/wp-includes/kses.php

@@ -964 +964 @@
  *
  * @since 1.0.0
+ * @since 6.6.0 Recognize additional forms of invalid HTML which convert into comments.
  *
  * @global array[]|string $pass_allowed_html      An array of allowed HTML elements and attributes,
@@ -982 +983 @@
     $pass_allowed_protocols = $allowed_protocols;
 
-    return preg_replace_callback( '%(<!--.*?(-->|$))|(<[^>]*(>|$)|>)%', '_wp_kses_split_callback', $content );
+    $token_pattern = <<<REGEX
+~
+    (                      # Detect comments of various flavors before attempting to find tags.
+        (<!--.*?(-->|$))   #  - Normative HTML comments.
+        |
+        </[^a-zA-Z][^>]*>  #  - Closing tags with invalid tag names.
+    )
+    |
+    (<[^>]*(>|$)|>)        # Tag-like spans of text.
+~x
+REGEX;
+    return preg_replace_callback( $token_pattern, '_wp_kses_split_callback', $content );
 }
 
@@ -1070 +1082 @@
  * @ignore
  * @since 1.0.0
+ * @since 6.6.0 Recognize additional forms of invalid HTML which convert into comments.
  *
  * @param string         $content           Content to filter.
@@ -1076 +1089 @@
  *                                          for the list of accepted context names.
  * @param string[]       $allowed_protocols Array of allowed URL protocols.
+ *
  * @return string Fixed HTML element
  */
@@ -1081 +1095 @@
     $content = wp_kses_stripslashes( $content );
 
-    // It matched a ">" character.
+    /*
+     * The regex pattern used to split HTML into chunks attempts
+     * to split on HTML token boundaries. This function should
+     * thus receive chunks that _either_ start with meaningful
+     * syntax tokens, like a tag `<div>` or a comment `<!-- ... -->`.
+     *
+     * If the first character of the `$content` chunk _isn't_ one
+     * of these syntax elements, which always starts with `<`, then
+     * the match had to be for the final alternation of `>`. In such
+     * case, it's probably standing on its own and could be encoded
+     * with a character reference to remove ambiguity.
+     *
+     * In other words, if this chunk isn't from a match of a syntax
+     * token, it's just a plaintext greater-than (`>`) sign.
+     */
     if ( ! str_starts_with( $content, '<' ) ) {
         return '&gt;';
     }
 
-    // Allow HTML comments.
+    /*
+     * When a closing tag appears with a name that isn't a valid tag name,
+     * it must be interpreted as an HTML comment. It extends until the
+     * first `>` character after the initial opening `</`.
+     *
+     * Preserve these comments and do not treat them like tags.
+     */
+    if ( 1 === preg_match( '~^</[^a-zA-Z][^>]*>$~', $content ) ) {
+        $content     = substr( $content, 2, -1 );
+        $transformed = null;
+
+        while ( $transformed !== $content ) {
+            $transformed = wp_kses( $content, $allowed_html, $allowed_protocols );
+            $content     = $transformed;
+        }
+
+        return "</{$transformed}>";
+    }
+
+    /*
+     * Normative HTML comments should be handled separately as their
+     * parsing rules differ from those for tags and text nodes.
+     */
     if ( str_starts_with( $content, '<!--' ) ) {
         $content = str_replace( array( '<!--', '-->' ), '', $content );

trunk/tests/phpunit/tests/kses.php

@@ -1933 +1933 @@
 
     /**
+     * Ensures that `wp_kses()` preserves various kinds of HTML comments, both valid and invalid.
+     *
+     * @ticket 61009
+     *
+     * @param string $html_comment    HTML containing a comment; must not be a valid comment
+     *                                but must be syntax which a browser interprets as a comment.
+     * @param string $expected_output How `wp_kses()` ought to transform the comment.
+     */
+    public function wp_kses_preserves_html_comments( $html_comment, $expected_output ) {
+        $this->assertSame(
+            $expected_output,
+            wp_kses( $html_comment, array() ),
+            'Failed to properly preserve HTML comment.'
+        );
+    }
+
+    /**
+     * Data provider.
+     *
+     * @return array[].
+     */
+    public static function data_html_containing_various_kinds_of_html_comments() {
+        return array(
+            'Normative HTML comment'            => array( 'before<!-- this is a comment -->after', 'before<!-- this is a comment -->after' ),
+            'Closing tag with invalid tag name' => array( 'before<//not a tag>after', 'before<//not a tag>after' ),
+        );
+    }
+
+    /**
      * Test that attributes with a list of allowed values are filtered correctly.
      *